I’m a software engineer building backend systems and AI / agentic tooling — the layer where models stop answering questions and start operating real infrastructure.
I work primarily in Go and Python, and I’m deliberately unattached to tools: ADK, LangChain, LangGraph, or plain HTTP calls and a state machine — whichever one actually fits the problem. Frameworks are a means. Reaching for the fashionable one costs more than writing the thing yourself.
What I care about is how a system behaves when things go wrong. Failure as a design input rather than a retrofit: timeouts and retries that are safe to repeat, tool access scoped to least privilege, and enough tracing that you can answer “what did the agent actually do?” after the fact. Agents that touch real systems need the same rigor as any distributed system — because that’s what they are.
Most of what I know about infrastructure I learned in public: contributing to librsvg with the GNOME Foundation, edge-case coverage for Kubernetes’ cloud-provider-openstack, and a PKCS#11 module for Namecoin.
When I’m not at the terminal — photography 📸, reading 📚, and the history of systems that shaped the modern world.
Available for backend and AI-systems engineering work — contract or full-time.